Add MongoDB Vector Search Tool

README.md

@@ -25,6 +25,7 @@ CrewAI provides an extensive collection of powerful tools ready to enhance your
 - **File Management**: `FileReadTool`, `FileWriteTool`
 - **Web Scraping**: `ScrapeWebsiteTool`, `SeleniumScrapingTool`
 - **Database Integrations**: `PGSearchTool`, `MySQLSearchTool`
+- **Vector Database Integrations**: `MongoDBVectorSearchTool`, `QdrantVectorSearchTool`, `WeaviateVectorSearchTool`
 - **API Integrations**: `SerperApiTool`, `EXASearchTool`
 - **AI-powered Tools**: `DallETool`, `VisionTool`, `StagehandTool`
 
@@ -226,4 +227,3 @@ Join our rapidly growing community and receive real-time support:
 - [Open an Issue](https://github.com/crewAIInc/crewAI/issues)
 
 Build smarter, faster, and more powerful AI solutions—powered by CrewAI Tools.
-

crewai_tools/__init__.py

@@ -1,5 +1,6 @@
 from .adapters.enterprise_adapter import EnterpriseActionTool
 from .adapters.mcp_adapter import MCPServerAdapter
+from .adapters.zapier_adapter import ZapierActionTool
 from .aws import (
     BedrockInvokeAgentTool,
     BedrockKBRetrieverTool,
@@ -23,9 +24,9 @@
     DirectorySearchTool,
     DOCXSearchTool,
     EXASearchTool,
+    FileCompressorTool,
     FileReadTool,
     FileWriterTool,
-    FileCompressorTool,
     FirecrawlCrawlWebsiteTool,
     FirecrawlScrapeWebsiteTool,
     FirecrawlSearchTool,
@@ -35,6 +36,8 @@
     LinkupSearchTool,
     LlamaIndexTool,
     MDXSearchTool,
+    MongoDBVectorSearchConfig,
+    MongoDBVectorSearchTool,
     MultiOnTool,
     MySQLSearchTool,
     NL2SQLTool,
@@ -76,4 +79,3 @@
     YoutubeVideoSearchTool,
     ZapierActionTools,
 )
-from .adapters.zapier_adapter import ZapierActionTool

crewai_tools/tools/__init__.py

@@ -16,10 +16,10 @@
 from .exa_tools.exa_search_tool import EXASearchTool
 from .file_read_tool.file_read_tool import FileReadTool
 from .file_writer_tool.file_writer_tool import FileWriterTool
+from .files_compressor_tool.files_compressor_tool import FileCompressorTool
 from .firecrawl_crawl_website_tool.firecrawl_crawl_website_tool import (
     FirecrawlCrawlWebsiteTool,
 )
-from .files_compressor_tool.files_compressor_tool import FileCompressorTool
 from .firecrawl_scrape_website_tool.firecrawl_scrape_website_tool import (
     FirecrawlScrapeWebsiteTool,
 )
@@ -30,6 +30,11 @@
 from .linkup.linkup_search_tool import LinkupSearchTool
 from .llamaindex_tool.llamaindex_tool import LlamaIndexTool
 from .mdx_search_tool.mdx_search_tool import MDXSearchTool
+from .mongodb_vector_search_tool import (
+    MongoDBToolSchema,
+    MongoDBVectorSearchConfig,
+    MongoDBVectorSearchTool,
+)
 from .multion_tool.multion_tool import MultiOnTool
 from .mysql_search_tool.mysql_search_tool import MySQLSearchTool
 from .nl2sql.nl2sql_tool import NL2SQLTool

crewai_tools/tools/mongodb_vector_search_tool/README.md

@@ -0,0 +1,87 @@
+# MongoDBVectorSearchTool
+
+## Description
+This tool is specifically crafted for conducting vector searches within docs within a MongoDB database. Use this tool to find semantically similar docs to a given query.
+
+MongoDB can act as a vector database that is used to store and query vector embeddings. You can follow the docs here:
+https://www.mongodb.com/docs/atlas/atlas-vector-search/vector-search-overview/
+
+## Installation
+Install the crewai_tools package with MongoDB support by executing the following command in your terminal:
+
+```shell
+pip install crewai-tools[mongodb]
+```
+
+or
+
+```
+uv add crewai-tools --extra mongodb
+```
+
+## Example
+To utilize the MongoDBVectorSearchTool for different use cases, follow these examples:
+
+```python
+from crewai_tools import MongoDBVectorSearchTool
+
+# To enable the tool to search any website the agent comes across or learns about during its operation
+tool = MongoDBVectorSearchTool(
+    database_name="example_database',
+    collection_name='example_collections',
+    connection_string="<your_mongodb_connection_string>",
+)
+```
+
+or
+
+```python
+from crewai_tools import MongoDBVectorSearchConfig, MongoDBVectorSearchTool
+
+# Setup custom embedding model and customize the parameters.
+query_config = MongoDBVectorSearchConfig(limit=10, oversampling_factor=2)
+tool = MongoDBVectorSearchTool(
+    database_name="example_database',
+    collection_name='example_collections',
+    connection_string="<your_mongodb_connection_string>",
+    query_config=query_config,
+    index_name="my_vector_index",
+    generative_model="gpt-4o-mini"
+)
+
+# Adding the tool to an agent
+rag_agent = Agent(
+    name="rag_agent",
+    role="You are a helpful assistant that can answer questions with the help of the MongoDBVectorSearchTool.",
+    goal="...",
+    backstory="...",
+    llm="gpt-4o-mini",
+    tools=[tool],
+)
+```
+
+Preloading the MongoDB database with documents:
+
+```python
+from crewai_tools import MongoDBVectorSearchTool
+
+# Generate the documents and add them to the MongoDB database
+test_docs = client.collections.get("example_collections")
+
+# Create the tool.
+tool = MongoDBVectorSearchTool(
+    database_name="example_database',
+    collection_name='example_collections',
+    connection_string="<your_mongodb_connection_string>",
+)
+
+# Add the text from a set of CrewAI knowledge documents.
+texts = []
+for d in os.listdir("knowledge"):
+    with open(os.path.join("knowledge", d), "r") as f:
+        texts.append(f.read())
+tool.add_texts(text)
+
+# Create the vector search index (if it wasn't already created in Atlas).
+tool.create_vector_search_index(dimensions=3072)
+```

crewai_tools/tools/mongodb_vector_search_tool/__init__.py

@@ -0,0 +1,11 @@
+from .vector_search import (
+    MongoDBToolSchema,
+    MongoDBVectorSearchConfig,
+    MongoDBVectorSearchTool,
+)
+
+__all__ = [
+    "MongoDBVectorSearchConfig",
+    "MongoDBVectorSearchTool",
+    "MongoDBToolSchema",
+]

crewai_tools/tools/mongodb_vector_search_tool/utils.py

@@ -0,0 +1,120 @@
+from __future__ import annotations
+
+from time import monotonic, sleep
+from typing import TYPE_CHECKING, Any, Callable, Dict, List, Optional
+
+if TYPE_CHECKING:
+    from pymongo.collection import Collection
+
+
+def _vector_search_index_definition(
+    dimensions: int,
+    path: str,
+    similarity: str,
+    filters: Optional[List[str]] = None,
+    **kwargs: Any,
+) -> Dict[str, Any]:
+    # https://www.mongodb.com/docs/atlas/atlas-vector-search/vector-search-type/
+    fields = [
+        {
+            "numDimensions": dimensions,
+            "path": path,
+            "similarity": similarity,
+            "type": "vector",
+        },
+    ]
+    if filters:
+        for field in filters:
+            fields.append({"type": "filter", "path": field})
+    definition = {"fields": fields}
+    definition.update(kwargs)
+    return definition
+
+
+def create_vector_search_index(
+    collection: Collection,
+    index_name: str,
+    dimensions: int,
+    path: str,
+    similarity: str,
+    filters: Optional[List[str]] = None,
+    *,
+    wait_until_complete: Optional[float] = None,
+    **kwargs: Any,
+) -> None:
+    """Experimental Utility function to create a vector search index
+
+    Args:
+        collection (Collection): MongoDB Collection
+        index_name (str): Name of Index
+        dimensions (int): Number of dimensions in embedding
+        path (str): field with vector embedding
+        similarity (str): The similarity score used for the index
+        filters (List[str]): Fields/paths to index to allow filtering in $vectorSearch
+        wait_until_complete (Optional[float]): If provided, number of seconds to wait
+            until search index is ready.
+        kwargs: Keyword arguments supplying any additional options to SearchIndexModel.
+    """
+    from pymongo.operations import SearchIndexModel
+
+    if collection.name not in collection.database.list_collection_names():
+        collection.database.create_collection(collection.name)
+
+    result = collection.create_search_index(
+        SearchIndexModel(
+            definition=_vector_search_index_definition(
+                dimensions=dimensions,
+                path=path,
+                similarity=similarity,
+                filters=filters,
+                **kwargs,
+            ),
+            name=index_name,
+            type="vectorSearch",
+        )
+    )
+
+    if wait_until_complete:
+        _wait_for_predicate(
+            predicate=lambda: _is_index_ready(collection, index_name),
+            err=f"{index_name=} did not complete in {wait_until_complete}!",
+            timeout=wait_until_complete,
+        )
+
+
+def _is_index_ready(collection: Collection, index_name: str) -> bool:
+    """Check for the index name in the list of available search indexes to see if the
+    specified index is of status READY
+
+    Args:
+        collection (Collection): MongoDB Collection to for the search indexes
+        index_name (str): Vector Search Index name
+
+    Returns:
+        bool : True if the index is present and READY false otherwise
+    """
+    for index in collection.list_search_indexes(index_name):
+        if index["status"] == "READY":
+            return True
+    return False
+
+
+def _wait_for_predicate(
+    predicate: Callable, err: str, timeout: float = 120, interval: float = 0.5
+) -> None:
+    """Generic to block until the predicate returns true
+
+    Args:
+        predicate (Callable[, bool]): A function that returns a boolean value
+        err (str): Error message to raise if nothing occurs
+        timeout (float, optional): Wait time for predicate. Defaults to TIMEOUT.
+        interval (float, optional): Interval to check predicate. Defaults to DELAY.
+
+    Raises:
+        TimeoutError: _description_
+    """
+    start = monotonic()
+    while not predicate():
+        if monotonic() - start > timeout:
+            raise TimeoutError(err)
+        sleep(interval)